Virtual tape

27.03.2006
Since the dawn of the digital computer age, long-term data storage and backup have been the province of a single primary medium: magnetic tape. Tape has compelling advantages. It's inexpensive to operate and buy, and even cheaper to store, whether it exists on reels, inside cartridges or as part of an automated tape library system. Tape also has the benefit of separating the portable and inexpensive storage medium from the larger, more costly recording machinery. The introduction of tape made it possible to back up everything, keep copies off-site and restore older or deleted files as needed.

In comparison, hard drive storage combined the machine and the medium into a single piece of hardware, gaining speed and simplifying access. But it also drove up storage costs and was for years simply not economical for backup use. Tape persisted as the storage medium of choice, even though it suffered from poor performance and the need for sequential, not random, access to stored data. It wasn't very fast, and for the most part, operations had to be run in batch modes, often overnight.

In the early 1970s, IBM predicted the death of tape as a backup medium, and since then, others (including Computerworld; see QuickLink 41762) have continued to echo that sentiment. That hasn't happened yet, and it's not at all obvious that it will. But the amount of data being stored and processed continues to grow exponentially, and while ever-larger tape formats continue to emerge, the time needed to perform regular backups is also growing.

Finally, the economics of backup changed radically as hard drive storage became far cheaper. Not only are new hard drives cheap, capacious, physically smaller and increasingly reliable, but they operate much faster and offer online storage at off-line prices -- and with no waiting. A 250GB hard drive today costs less per gigabyte than the digital linear tape cartridges for a relatively recent tape library. Although tapes are still much more portable than RAID arrays, it's now practical to replace tape with disk for primary backups to boost speed, improve reliability and eliminate delays in loading and searching for needed data.

One logical response to this technological change was for enterprise IT to shift to hard-drive-based backup systems. But this approach required a surprising amount of work to convert existing systems, policies and procedures. Enterprise backup teams are used to fine-tuning backup environments and applications by adding custom scripts and workflows to manage thousands of individual tapes both on- and off-site. Even positive change will be disruptive in this setting, so IT managers are rightly concerned about the effects of disk-based backups on their systems and scheduling. The better answer, at least for now, turns out to be a game of "Let's Pretend."

With virtual tape, even though we're backing up direct to disk, we pretend we're dealing with tape. Data is backed up to the disk subsystem by accessing it through what's called a virtual tape library -- software that emulates the properties of tape. By making the disks look like tape, the virtual system lets IT use its existing tape-based scheduling procedures and practices, scripts and workflows; the only difference is that backup data is stored on a different set of devices. This is such a simplification as to be nearly simple-minded, but it allows IT to expand its capabilities with little or no effort, and it gets away from the need to handle, rotate and store near-line tapes. The net effect is that virtual tape makes both backups and restores faster, more reliable and cheaper.